bytedance/deer-flow

An open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.

What it solves

DeerFlow is a "super agent harness" designed to orchestrate complex tasks by coordinating sub-agents, memory systems, and secure sandboxes. It enables the creation of autonomous agents capable of deep research and general-purpose task execution through extensible skills.

How it works

DeerFlow operates as a central orchestrator that manages several key components:

  • Sub-Agents: It can spawn and manage specialized agents to handle specific parts of a larger goal.
  • Memory: It utilizes long-term memory and context engineering (including manual context compaction) to maintain coherence over long sessions.
  • Sandboxes: It provides a secure environment and file system for agents to execute code and manipulate files safely.
  • Skills & Tools: It integrates with various tools, including web search (via InfoQuest, Jina, Browserless) and coding tools like Claude Code.
  • Model Flexibility: It supports a wide range of LLM providers (OpenAI, DeepSeek, vLLM, etc.) and allows for fine-tuned configuration of reasoning and thinking modes.

Who it’s for

It is intended for developers and researchers who need a powerful framework to build and deploy autonomous AI agents that can perform deep research, execute code in isolated environments, and manage complex, multi-step workflows.

Highlights

  • Super Agent Architecture: Orchestrates multiple sub-agents to tackle complex problems.
  • Secure Execution: Integrated sandbox and file system for safe tool use.
  • Extensible Skills: Easy integration of new tools and capabilities.
  • Advanced Memory: Features long-term memory and context compaction to handle large amounts of information.
  • Broad Model Support: Native adapters for various providers, including specialized support for DeepSeek's reasoning content.
  • Developer Tooling: Includes a TUI (Terminal Workbench) and integrates with tracing tools like LangSmith, Langfuse, and Monocle.

Related

  • Project
    withastro/flue

    A programmable TypeScript harness for building autonomous agents with built-in support for secure sandboxes, durable sessions, and tool integration.

  • Project
    MiroMindAI/MiroFlow

    MiroFlow is an open-source research agent framework for multi-step internet research and complex reasoning tasks, such as future event prediction.

  • Project
    truefoundry/trueforge

    TrueForge is an open‑source TypeScript runtime that executes LLM agents. It manages model calls, tool invocations, sandboxed code, approvals, and session state, exposing a chat UI, HTTP API, and TypeScript SDK. Deploy locally (SQLite) or in hosted mode (Postgres + Redis) via Docker‑Compose, Helm, or Railway.

  • Project
    alikon-art/DeterminFlow

    A production-oriented AI workflow framework that organizes LLMs, scripts, and APIs into versioned, recoverable nodes to ensure stable and cost-effective execution.

  • Project
    langchain-ai/deepagents

    An opinionated, batteries-included agent harness built on LangGraph that provides out-of-the-box planning, context management, and tool integration for long-horizon tasks.